CN-121984069-A - Electric automobile and power grid joint scheduling method, device, equipment and storage medium

CN121984069ACN 121984069 ACN121984069 ACN 121984069ACN-121984069-A

Abstract

The application provides a method, a device, equipment and a storage medium for joint scheduling of an electric automobile and a power grid, and relates to the technical field of artificial intelligence. The method comprises the steps of encoding a joint scheduling problem of an electric automobile and a power grid into a two-part graph structure, modeling the joint scheduling problem of the electric automobile and the power grid into MISOCP, wherein MISOCP is limited by target constraint, the target constraint comprises power distribution network constraint and vehicle network cooperative constraint, scheduling prediction is carried out according to the two-part graph structure to obtain scheduling prediction probability, a confidence domain is constructed according to the scheduling prediction probability, and solving sub-problems determined by the confidence domain in the confidence domain with the aim of minimizing power generation cost and electric automobile energy interaction cost to obtain scheduling parameters in a scheduling period. According to the scheme, the dispatching parameters are solved in the confidence domain, so that the calculation complexity of joint dispatching of the electric automobile and the power grid can be reduced, and the calculation cost is reduced.

Inventors

YAO XI
ZHOU WEI
TANG SHIQI
HUANG JUAN
LI TENGDA
WANG YU

Assignees

中移(上海)信息通信科技有限公司
中移智行网络科技有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (11)

1. The electric automobile and power grid joint scheduling method is characterized by comprising the following steps of: encoding the electric vehicle and power grid joint scheduling problem into a two-part graph structure, wherein the electric vehicle and power grid joint scheduling problem is modeled into a mixed integer second order cone plan MISOCP, and MISOCP is limited by target constraints, wherein the target constraints comprise power distribution network constraints and vehicle network cooperative constraints; According to the two-part graph structure, scheduling prediction is carried out, and scheduling prediction probability is obtained; constructing a confidence domain according to the scheduling prediction probability; solving the sub-problems determined by the confidence domain in the confidence domain with the aim of minimizing the power generation cost and the energy interaction cost of the electric automobile as well as obtaining the scheduling parameters in the scheduling period; the sub-problem is composed of all target constraints and confidence domain constraints of the electric automobile and power grid joint scheduling problem.
2. The method of claim 1, wherein the scheduling parameters include at least one of generator power, generator voltage, electric vehicle charge capacity, electric vehicle discharge capacity, electric vehicle residual capacity, electric vehicle charge power, electric vehicle discharge power, and electric vehicle charge-discharge mode.
3. The method of claim 1, wherein the encoding the electric vehicle and grid joint scheduling problem into a bipartite graph structure comprises: constructing a variable node set and a constraint node set based on the power grid physical nodes and the electric vehicle aggregation unit; mapping continuous variables in optimal power flow calculation and integer variables in electric vehicle dispatching into node attributes of two graphs, and representing target constraint as cross-set edge connection to obtain a two graph structure with nodes and edges; The integer variable is used for representing the charge and discharge states of the electric automobile in each scheduling period, and the continuous variable is used for describing at least one of tidal current power, voltage amplitude and residual electric quantity level.
4. The method of claim 1, wherein performing scheduling prediction according to the bipartite graph structure to obtain scheduling prediction probability comprises: embedding, representing and learning the two-part graph structure by adopting a graph neural network GCNN, and extracting space structure information between variables and constraints; And according to the space structure information, the scheduling prediction probability of each time step is obtained by utilizing the dynamic scheduling behavior of the long-short-term memory network LSTM processing time dimension.
5. The method of claim 4, wherein the embedding representation learning of the bipartite graph structure using the graph neural network GCNN, extracting spatial structure information between variables and constraints, comprises: Aggregating information of variable nodes and edges adjacent to a first constraint node in a bipartite graph structure, and updating embedded representation of the first constraint node; Aggregating information of adjacent constraint nodes and edges of a first variable node in a two-part graph structure, and updating embedded representation of the first variable node; Determining spatial structure information between the variable and the constraint based on the embedded representation of the first constraint node and the embedded representation of the first variable node; the first constraint node is any node in a constraint node set corresponding to the bipartite graph structure, and the first variable node is any node in a variable node set corresponding to the bipartite graph structure.
6. The method of claim 4, wherein the obtaining the scheduling prediction probability of each time step according to the spatial structure information by using the dynamic scheduling behavior of the long-short-term memory network LSTM to process the time dimension comprises: And (3) incrementally inputting the control variable of each electric vehicle in each time period and the space structure information corresponding to the node of the charging station in each time period into an LSTM model according to time, capturing the inter-time period state evolution and constraint transfer of the electric vehicles, and obtaining the scheduling prediction probability of each time step.
7. The method of claim 1, wherein constructing a confidence domain based on the scheduling prediction probabilities comprises: Obtaining k positions with maximum prediction probability in all scheduling prediction probabilities of each time step, wherein k is an integer greater than or equal to 1; constructing an initial solution according to the modes corresponding to the k positions, wherein the mode corresponding to each position is one of a charging mode, a discharging mode and an idle mode; and determining a confidence radius according to the initial solution by using a target norm, and acquiring a confidence domain.
8. The method of claim 1, wherein the distribution network constraints include at least one of: Grid power flow constraint, power balance constraint, voltage amplitude constraint, current amplitude constraint, equipment capacity and operation limit constraint; And/or The vehicle network cooperative constraint comprises at least one of the following: Charge-discharge state constraints, charge-discharge power constraints, and battery capacity constraints.
9. An electric automobile and electric wire netting joint scheduling device, characterized by comprising: the coding module is used for coding the electric automobile and power grid joint scheduling problem into a two-part graph structure, the electric automobile and power grid joint scheduling problem is modeled into a mixed integer second order cone plan MISOCP, the MISOCP is limited by target constraint, and the target constraint comprises power distribution network constraint and vehicle network cooperative constraint; The first acquisition module is used for carrying out scheduling prediction according to the bipartite graph structure to acquire scheduling prediction probability; the construction module is used for constructing a confidence domain according to the scheduling prediction probability; The second acquisition module is used for solving the sub-problems determined by the confidence domain in the confidence domain with the aim of minimizing the power generation cost and the energy interaction cost of the electric automobile as well as acquiring the scheduling parameters in the scheduling period; the sub-problem is composed of all target constraints and confidence domain constraints of the electric automobile and power grid joint scheduling problem.
10. The electric automobile and power grid joint scheduling device comprises a transceiver, a processor, a memory and a program or an instruction stored on the memory and capable of running on the processor, and is characterized in that the electric automobile and power grid joint scheduling method according to any one of claims 1-8 is realized when the processor executes the program or the instruction.
11. A readable storage medium having stored thereon a program or instructions, which when executed by a processor, realizes the steps in the electric vehicle and grid joint scheduling method according to any one of claims 1-8.

Description

Electric automobile and power grid joint scheduling method, device, equipment and storage medium Technical Field The application relates to the technical field of artificial intelligence, in particular to a method, a device, equipment and a storage medium for joint scheduling of an electric automobile and a power grid. Background With the large-scale popularization of electric vehicles (ELECTRIC VEHICLE, EV) and the continuous pushing of clean energy policies, the unified dispatching control demands of the electric network on the charging and discharging behaviors of the electric vehicles are increasing. The conventional electric automobile dispatching generally processes the optimal power flow (Optimal Power Flow, OPF) and the Vehicle-to-Grid (V2G) behavior separately, and adopts a mixed integer programming model to jointly model and optimize the charge-discharge strategy and the Grid running state of the EV. The scheme takes the power flow balance of the power system as a core, and solves the problem of minimizing the system cost through global optimization under the physical constraint conditions of node voltage, power limit, state of Charge (SOC) boundary and the like. In this process, the state of the EV is modeled by Integer variables, creating a typical Mixed-Integer Second order cone programming (Mixed-Inter Second-Order Cone Programming, MISOCP) problem. However, due to the large scale, multiple variables and strong nonlinearity of the OPF-V2G problem, the conventional solver (such as solving constraint integer programming (Solving Constraint Integer Programs, SCIP)) needs to traverse a large amount of state space in the solving process, so that the computing overhead is extremely high, and the real-time requirement is difficult to meet. Disclosure of Invention The application aims to provide a method, a device, equipment and a storage medium for joint scheduling of an electric automobile and a power grid, which solve the problems of high calculation complexity and high calculation cost in joint scheduling of the electric automobile and the power grid. To achieve the above objective, an embodiment of the present application provides a method for jointly dispatching an electric vehicle and a power grid, including: encoding the electric vehicle and power grid joint scheduling problem into a two-part graph structure, wherein the electric vehicle and power grid joint scheduling problem is modeled into a mixed integer second order cone plan MISOCP, and MISOCP is limited by target constraints, wherein the target constraints comprise power distribution network constraints and vehicle network cooperative constraints; According to the two-part graph structure, scheduling prediction is carried out, and scheduling prediction probability is obtained; constructing a confidence domain according to the scheduling prediction probability; solving the sub-problems determined by the confidence domain in the confidence domain with the aim of minimizing the power generation cost and the energy interaction cost of the electric automobile as well as obtaining the scheduling parameters in the scheduling period; the sub-problem is composed of all target constraints and confidence domain constraints of the electric automobile and power grid joint scheduling problem. Optionally, the dispatching parameters comprise at least one of generator power, generator voltage, electric vehicle charging energy, electric vehicle discharging energy, electric vehicle remaining energy, electric vehicle charging power, electric vehicle discharging power and electric vehicle charging and discharging modes. Optionally, the encoding the electric automobile and power grid joint scheduling problem into a bipartite graph structure includes: constructing a variable node set and a constraint node set based on the power grid physical nodes and the electric vehicle aggregation unit; mapping continuous variables in optimal power flow calculation and integer variables in electric vehicle dispatching into node attributes of two graphs, and representing target constraint as cross-set edge connection to obtain a two graph structure with nodes and edges; The integer variable is used for representing the charge and discharge states of the electric automobile in each scheduling period, and the continuous variable is used for describing at least one of tidal current power, voltage amplitude and residual electric quantity level. Optionally, the performing scheduling prediction according to the bipartite graph structure, to obtain scheduling prediction probability, includes: embedding, representing and learning the two-part graph structure by adopting a graph neural network GCNN, and extracting space structure information between variables and constraints; And according to the space structure information, the scheduling prediction probability of each time step is obtained by utilizing the dynamic scheduling behavior of the long-short-term memory network LSTM proc